/haojieliu_PLIN_Based_on_3D_Motion_Representation_and_Spatial_Super

Pseudo-lidar point cloud interpolation based on 3d motion representation and spatial supervision

Primary LanguagePython

paper

Pseudo LiDAR Point Cloud Interpolation Based on 3D Motion Representation and Spatial Supervision

Notes

Our network is trained with the KITTI dataset alone, without pretraining on Cityscapes or other similar driving dataset (either synthetic or real).

Requirements

This code was tested with Python 3 and PyTorch 1.0 on Ubuntu 16.04,chamferdist.

  • Install PyTorch on a machine with CUDA GPU.

  • The code for self-supervised training requires OpenCV along with the contrib modules. For instance,

  • Download the KITTI Depth Dataset and the corresponding RGB images.

  • The code, data and result directory structure is shown as follows


├── data
|   ├── kitti_depth
|   |   ├── train
|   |   ├── val_selection_cropped
|   └── kitti_rgb
|   |   ├── train
|   |   ├── val_selection_cropped
├── results

Training and testing

A complete list of training options is available with

python main.py -h

For instance,

python main.py --train-mode dense -b 1 # train with the KITTI semi-dense annotations and batch size 1
python main.py --resume [checkpoint-path] # resume previous training
python main.py --evaluate [checkpoint-path] # test the trained model
eg.
CUDA_VISIBLE_DEVICES=1 python main.py --train-mode dense -b 1
python main.py --evaluate ./model_best.pth.tar

Questions

Please create a new issue for code-related questions. Pull requests are welcome.

Citation

If you use our code or method in your work, please cite the following:

@article{liu2021pseudo,
	title={Pseudo-LiDAR Point Cloud Interpolation Based on 3D Motion Representation and Spatial Supervision},
	author={Liu, Haojie and Liao, Kang and Lin, Chunyu and Zhao, Yao and Guo, Yulan},
	journal={IEEE Transactions on Intelligent Transportation Systems},
	year={2021},
	publisher={IEEE}
	}